CN116861300B - Automatic auxiliary labeling method and device for complex maneuvering data set of military machine type - Google Patents
Automatic auxiliary labeling method and device for complex maneuvering data set of military machine type Download PDFInfo
- Publication number
- CN116861300B CN116861300B CN202311116892.7A CN202311116892A CN116861300B CN 116861300 B CN116861300 B CN 116861300B CN 202311116892 A CN202311116892 A CN 202311116892A CN 116861300 B CN116861300 B CN 116861300B
- Authority
- CN
- China
- Prior art keywords
- sequence
- maneuver
- flight parameter
- flight
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000002372 labelling Methods 0.000 title claims abstract description 36
- 238000000034 method Methods 0.000 claims abstract description 33
- 238000007781 pre-processing Methods 0.000 claims abstract description 21
- 239000011159 matrix material Substances 0.000 claims abstract description 17
- 238000012800 visualization Methods 0.000 claims abstract description 17
- 230000009471 action Effects 0.000 claims description 78
- 238000012549 training Methods 0.000 claims description 15
- 230000001133 acceleration Effects 0.000 claims description 8
- 239000013598 vector Substances 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 description 6
- 239000012634 fragment Substances 0.000 description 5
- 238000012545 processing Methods 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 3
- 230000009189 diving Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000010921 in-depth analysis Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000001303 quality assessment method Methods 0.000 description 1
- 238000013441 quality evaluation Methods 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000005303 weighing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to an automatic auxiliary labeling method and device for a complex maneuvering data set of a military machine type. The method comprises the following steps: collecting historical flight parameter sequence data and selecting a standard maneuver sample sequence; preprocessing historical flight parameter sequence and target maneuver template sequence data; using a Matrix Profile data structure, and using a target maneuvering template sequence to preliminarily match in a historical flight parameter sequence and extracting a to-be-identified flight parameter subsequence; pre-classifying the identification subsequence; performing similarity matching on the flight parameter subsequence to be identified and the target maneuver template sequence by adopting a multidimensional dynamic time planning method, and finishing the fine classification of maneuver; and carrying out three-dimensional visualization on the flight parameter subsequence, and manually rechecking the automatically marked maneuver categories to obtain a maneuver data set with the tag. The method can remarkably improve the labeling efficiency and accuracy of the complex maneuvering data set, and is suitable for various military models.
Description
Technical Field
The invention belongs to the technical field of intelligent flight data processing, and particularly relates to an automatic auxiliary labeling method and device for a military model complex maneuver data set.
Background
Scientific evaluation of flight training quality is of great importance for analyzing pilot maneuvering habits and improving pilot flight driving technologies and ensuring flight safety. Maneuver identification is a key step in flight training quality assessment, and a large amount of assessment content is based on the acquisition of a specific category of action sequences. The flight parameter data recorded and stored by the airborne flight parameter system in the form of multi-dimensional time sequence data provides objective and scientific basis for motor action recognition.
Currently, the maneuver identification method based on the flight parameter data mainly comprises a method based on pattern matching, a method based on an expert system and an intelligent method based on deep learning. Conventional pattern matching based methods require manual adjustment of the threshold. The expert system-based method requires the establishment of a manual rule knowledge base according to prior knowledge of domain experts. These two types of methods are to be further improved for complex maneuver recognition accuracy with high similarity. In recent years, artificial intelligence has been rapidly developed and advanced, and deep learning-based methods have shown remarkable performance in the field of motor action recognition. However, deep learning based methods require a large number of labeled maneuver sample data sets for model training. In addition, the flying parameter data recorded by the multimode onboard sensor such as a gyroscope or an accelerometer contains tens to hundreds of parameters, which is far more difficult to understand than the data of the other modal sensor such as a camera. The traditional maneuver data set labeling method is to observe and analyze flight data manually, identify maneuver fragments in the flight data and label the maneuver fragments. However, this manual labeling method has the following problems: (1) The workload is large, and a large amount of time and manpower resources are consumed for manually marking a large amount of flight data; (2) The subjectivity is strong, and subjective differences can exist in the identification and labeling of different personnel on the maneuver, so that the results are inconsistent; (3) The accuracy is low, the extraction and the category labeling of the maneuvering sequence fragments require professional staff, the labeling accuracy is influenced by factors such as individual ability, experience, fatigue and the like, and the labeling error can exist. Therefore, there is a need for a method and apparatus that automatically assists in labeling maneuvers in flight data to improve labeling efficiency and accuracy.
Disclosure of Invention
The technical problem to be solved by the invention is to provide an automatic auxiliary labeling technical scheme for a military machine type complex maneuver data set aiming at the defects of the prior art, so as to solve the technical problems of low labeling precision and low efficiency of target maneuver types under massive flight parameter data.
In order to achieve the aim of the invention, the invention adopts the following technical scheme:
s1, acquiring historical flight parameter sequence data of pilot flight training, and selecting a standard maneuver sample sequence as a target maneuver template;
s2, preprocessing the collected historical flight parameter sequence data and the selected target maneuver template sequence;
s3, based on a Matrix Profile data structure, using a target maneuvering template sequence to preliminarily match in a historical flight parameter sequence subjected to data preprocessing and extracting a flight parameter subsequence to be identified;
s4, pre-classifying the subsequence to be identified extracted in the S3 through obvious features, namely pre-matching the maneuver categories;
s5, performing similarity matching on the to-be-identified flight parameter subsequence and the target maneuver template sequence in the pre-classified maneuver sequence by adopting an MDTW algorithm, and completing fine classification of maneuver;
and S6, carrying out three-dimensional visualization on the sub-sequences of the flight parameters subjected to the fine classification, and carrying out manual accurate rechecking on the automatically marked maneuver class labels to obtain a final marked maneuver data set.
Preferably, in S1, historical flight parameter sequence data of flight training of the pilot is collected, and a standard maneuver sample sequence is selected as a target maneuver template; the method specifically comprises the following steps:
s1-1, acquiring historical flight parameter sequence data of pilot flight training, and decoding the historical flight parameter data into table data which can be directly read by a computer;
s1-2, determining the type of the target maneuver to be marked according to the flight training outline and the actual demand, and selecting a flight parameter subsequence meeting maneuver operation specification requirements in the flight training manual as a standard target maneuver template sequence.
Preferably, in S2, data preprocessing is carried out on the collected historical flight parameter sequence and the selected target maneuver template sequence; the method specifically comprises the following steps:
s2-1, acquiring parameters of pitch angle, inclination angle, air pressure height, X-axis angular velocity, Y-axis angular velocity, Z-axis angular velocity, horizontal acceleration and vertical acceleration which can judge different action types, filling missing values in a historical flight parameter sequence and a target maneuvering template sequence, and carrying out standardized preprocessing on each dimension in original data.
The data normalization processing formula is as follows:wherein (1)>Representing the mean value of each dimension of the flight data, < >>Representing the standard deviation of each dimension of data in the flight data. The average value of each dimension of data after pretreatment is 0, and the standard deviation is 1.
Preferably, in the step S3, based on a Matrix Profile data structure, a target maneuvering template sequence is used for preliminary matching in a historical flight parameter sequence subjected to data preprocessing, and a to-be-identified flight parameter subsequence is extracted; the method specifically comprises the following steps: given a time series of data-preprocessed historical flight parametersTarget maneuver template sequence for querying +.>. Use of the target maneuver template sequence +.>Sliding window of length from historical flight parameter time series subjected to data preprocessing +.>Starting to slide, calculating the sub-sequence within the window and the target maneuver template sequence each time +.>Is a distance of ∈length of generation ∈>Is less than a threshold value +.>The value of (2) is at the position from +.>The sub-sequence of the flight parameter to be identified which is matched and extracted in the step (a)>。
Wherein,representing the number of parameters>Representing the length of the time series of historical flight parameters, +.>Representing the length of the target maneuver template sequence, typically +.>Far less than->。
Preferably, in S4, the sub-sequence to be identified extracted in S3 is subjected to pre-classification processing, that is, pre-matching of maneuver categories is performed; the method specifically comprises the following steps:
s4-1, the obvious characteristics are height and pitch angle, and the pre-classified categories comprise a bucket like a jin, a jump like a dive, a lift turn like a sharp turn, a roll like a roll and a spiral like four categories;
s4-2, pre-matching of the maneuvering action category is specifically as follows: first, a threshold value is setAnd threshold->，/>Taking the average of the maximum height and minimum height differences in each standard target maneuver template sequence, +.>Taking the average value of the difference values of the maximum pitch angle and the minimum pitch angle in each standard target maneuvering template sequence, and then, according to the average value of the absolute value of the first-order difference of the heights of the subsequences to be identified, judging whether the average value is larger than a threshold value>Is divided into lifting actions, which are smaller than a threshold +.>Is classified into a non-lifting type of motion. In the lifting action, according to the average value of the absolute value of the first-order difference of the pitch angle of the subsequence to be identified, the average value is larger than a threshold value +.>The sequence of (1) is divided into a bucket-like body and a diving jump-like body, which is smaller than a threshold value +.>The sequence of (1) is divided into lifting turns; in the non-lifting type action, according to the average value of the absolute value of the first-order difference of the pitch angle of the subsequence to be identified, the average value is larger than a threshold value +.>Is divided into classes of convolutions, the mean value is smaller than the threshold value +.>Is divided into a sharp turn-like and a roll-like sequence;
preferably, in the maneuver identification stage in S5, similarity matching is carried out on the flight parameter subsequence to be identified and the target maneuver template sequence by utilizing an MDTW algorithm, so as to finish the fine classification of maneuver; the method specifically comprises the following steps:
s5-1, performing similarity matching on the to-be-identified flight parameter subsequence obtained after the pre-classification in the step S4 and the existing target maneuver template sequences, and obtaining a similarity value sequence by respectively calculating MDTW distances between the to-be-identified flight parameter subsequence and the C target maneuver template sequences，/>A value less than a threshold valueThe action to be recognized is determined as +.>A corresponding standard action category. Wherein->And setting according to the MDTW distance value of the actual maneuvering sequence and the corresponding target maneuvering template sequence.
S5-2, calculating similarity values of the two action sequences by adopting the MDTW distance calculating principle. Assume that action sequence 1 isAction sequence 2 is->MDTW path matrix is +.>，/>The dimensions of the sequence of actions are represented,and->Indicates the length of action sequence 1 and action sequence 2, respectively,/->Weights representing the respective dimensions;
the action sequences 1 and 2 are defined as: wherein (1)>Is the +.1 of action sequence>Dimension features at%>The value of the individual points,/->Is the +.2 of action sequence>Dimension features at%>The value of each point.
The MDTW path matrix is defined as: ,/>wherein, the method comprises the steps of, wherein,is +.1 of action sequence>Dimension features at%>The value of the individual points,/->Is +.2 for action sequence>Dimension features at%>The value of each point. />Representation->The weight of the dimension is determined by the weight of the dimension,/>representing action sequence 1->Value vector of all dimension characteristics of each point and action sequence 2 +.>Frame weighted matching distance of value vectors of all dimension characteristics of each point;
s5-3, multidimensional dynamic time planning distance of action sequence 1 and action sequence 2Defined as the accumulated distance of the optimal regular path +.>And the regular path meets three constraint conditions of boundary, continuity and monotonicity, and the formula is defined as follows:
wherein the boundary condition is that the start point of the regular path must be +.>The termination point must be +.>The method comprises the steps of carrying out a first treatment on the surface of the The continuity condition is that the regular path should be continuous, i.e. jump from one point to the next, either right, up or up right; the monotonicity condition is that the direction of movement of the regular path should be monotonic, i.e. not move in the opposite direction. Constructing an optimal regular path according to the three constraint conditions, and calculating the accumulated distance of the optimal regular path to obtain +.>。
Preferably, S6, carrying out three-dimensional visualization on the sub-sequences of the flight parameters subjected to fine classification, and rechecking the automatically marked maneuver class labels to obtain a final marked maneuver data set; the method specifically comprises the following steps:
s6-1, carrying out three-dimensional visualization on flight attitude by using pitch angle, roll angle and yaw angle information on the flight parameter subsequence automatically and finely classified by the MDTW algorithm, carrying out three-dimensional visualization on flight track by using altitude, longitude and latitude information on the flight parameter subsequence automatically and finely classified by the MDTW algorithm, rechecking the automatically marked maneuver type label according to the three-dimensional visualization result of the flight attitude and track, storing the flight parameter subsequence and the corresponding label type when the classification result of the MDTW algorithm is consistent with the manual classification result, taking the manual classification result as the type label when the classification result of the MDTW algorithm is inconsistent with the manual classification result, and storing the corresponding flight parameter subsequence, thus obtaining the maneuver data set with the label.
The invention also provides an auxiliary labeling device for the complex maneuvering data set of the military machine type.
Specifically, a military machine type complex maneuver data set auxiliary labeling device comprises:
the data acquisition module is used for acquiring the flight parameter sequence data;
the preprocessing module is used for carrying out data preprocessing on the flight parameter sequence data to generate preprocessed flight parameter sequence data;
the pre-matching module is used for carrying out preliminary matching on the target action template and the historical flight parameter sequence to obtain a flight parameter subsequence to be identified
And the maneuver classifying module is used for carrying out similarity matching calculation on the to-be-identified flight parameter subsequence and the target maneuver template sequence to obtain a fine classification result of maneuver.
And the maneuvering motion three-dimensional visualization module is used for carrying out three-dimensional visualization on the flight gesture and the track of the maneuvering motion sequence subjected to fine classification, and further carrying out manual accurate rechecking on the maneuvering type automatically marked.
The technical scheme provided by the invention has at least the following beneficial effects:
1. according to the invention, the Matrix Profile data structure is utilized, and the maneuvering action templates are used for automatically matching and extracting the target maneuvering action subsequences in the massive flight parameter data, so that the positioning of the time period where the target maneuvering action is positioned in the massive historical flight parameters and the extraction efficiency of the subsequence fragments are greatly improved, and the workload of labeling personnel is reduced.
2. According to the invention, after the sub-sequences of the flight parameters to be identified are pre-classified, the MDTW algorithm is adopted to calculate the similarity between the multi-dimensional and unequal-length maneuvering action sequences, so that the precise automatic identification of maneuvering types of the sub-sequences of the flight parameters to be identified is realized, the calculation amount of the MDTW is reduced, and the identification precision and efficiency are improved.
3. The automatic auxiliary labeling method for the motor actions of the military machine model is a universal method, can obviously improve the labeling efficiency of a complex motor action data set, is suitable for various different military machine models, and has wide engineering application prospects.
In summary, the automatic auxiliary labeling method and device for the maneuver data set provided by the invention realize automatic auxiliary labeling of maneuver fragments in the flight data by combining the Matrix Profile data structure and the MDTW algorithm, have the advantages of automation, accuracy, high efficiency and universality, and can be widely applied to flight action analysis and quality evaluation in the military aviation field.
Drawings
FIG. 1 is a general flow chart of a method for automatically assisting in labeling a complex maneuver data set for a military machine in accordance with one embodiment of the present invention.
FIG. 2 is a flow chart of a maneuver pre-classification process according to an embodiment of the present invention.
Fig. 3 is a schematic diagram of an MDTW trace-regular path according to an embodiment of the present invention.
FIG. 4 is a flowchart of an embodiment of the invention for automatically assisted labeling of a maneuver data set.
Fig. 5 is a schematic structural diagram of an automatic auxiliary labeling device for a complex maneuver data set of a military machine according to an embodiment of the present invention.
Detailed Description
Examples
For the purposes, technical solutions and advantages of the present application, the technical solutions of the present application will be clearly and completely described below with reference to specific embodiments of the present application and corresponding drawings. It will be apparent that the described embodiments are only some, but not all, of the embodiments of the present application. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure.
As shown in fig. 1, the embodiment of the invention provides an automatic auxiliary labeling method for a complex maneuvering data set of a military machine type, which comprises the following steps:
s1, collecting historical flight parameter sequence data of pilot flight training, and selecting a standard maneuver sample sequence as a target maneuver template.
During the flight of a military model, the onboard multisensor monitors a series of parameters related to the flight state. These flight state related parameters include pitch angle, roll angle, heading angle, angle of attack and sideslip angle, altitude, speed of flight, angular velocity, acceleration, etc. Thus, during annotation of maneuvers, different maneuvers may be described by in-depth analysis of the maneuver's flight parameter variation characteristics. It should be emphasized that these specific index parameters do not limit the scope of protection of the present application. In addition, all the maneuvering data in the complete flight time period from the start of the military model to the end of the flight can be acquired, and the flight parameter sequence data of a part of the complete flight time period of the military model can be selectively acquired. It should be clear that the time period for acquiring the flight parameter sequence data does not limit the scope of protection of the present application.
Further, a standard flight parameter subsequence is selected as a target maneuver template according to the operation specification of the target maneuver in the flight training manual. It is emphasized that in a preferred embodiment provided herein, the historical flight parameter sequence and standard maneuver sample sequence data obtained includes flight state related parameters such as pitch angle, roll angle, heading angle, angle of attack and sideslip angle, altitude, speed of flight, angular velocity, acceleration, and the like. Furthermore, the number of target maneuver templates is equal to the number of labeling categories of maneuver, which is not limiting to the scope of the present application.
S2, preprocessing data of the collected historical flight parameter sequence and the selected target maneuver template sequence.
Redundant information exists in the flight state parameters included in the acquired original historical flight parameter sequence and the target maneuver template sequence, and when maneuver judgment is carried out, the calculation efficiency is improved in order to reduce the data dimension, and the key parameters with the maneuver distinction degree are selected.
Further, in a preferred embodiment provided in the present application, pitch angle, tilt angle, barometric altitude, X-axis angular velocity, Y-axis angular velocity, Z-axis angular velocity, horizontal acceleration, and vertical acceleration parameters that can be determined for different motion classes are taken, missing values in the historical flight parameter sequence and the target maneuver template sequence are filled, and then standardized preprocessing is performed for each dimension in the raw data.
The data normalization processing formula is as follows:
wherein,representing the mean value of each dimension of the flight data, < >>Representing the standard deviation of each dimension of data in the flight data. The average value of each dimension of data after pretreatment is 0, and the standard deviation is 1.
And S3, based on a Matrix Profile data structure, using a target maneuvering template sequence to preliminarily match in a historical flight parameter sequence subjected to data preprocessing, and extracting a flight parameter subsequence to be identified.
Given a time series of data-preprocessed historical flight parametersTarget maneuver template sequence for querying +.>. Using a target maneuver template sequenceSliding window of length from historical flight parameter time series subjected to data preprocessing +.>Starting to slide, calculating the sub-sequence within the window and the target maneuver template sequence each time +.>Is a distance of ∈length of generation ∈>Is less than a threshold value +.>The value of (2) is at the position from +.>The sub-sequence of the flight parameter to be identified which is matched and extracted in the step (a)>。
Wherein,representing the number of parameters>Representing the length of the time series of historical flight parameters, +.>Representing the length of the target maneuver template sequence, typically +.>Far less than->Threshold->Is set manually.
S4, pre-classifying the subsequence to be identified extracted in the S3 through obvious features, namely pre-matching the maneuver categories.
The pre-classified categories comprise four categories of a weighing bucket, a diving jump, a lifting turn, a sharp turn, a rolling turn and a spiral turn. It will be appreciated that the pre-categorized categories described herein are not obviously limiting to the specific scope of protection of the present application.
Further, as shown in fig. 2, in a preferred embodiment provided in the present application, the maneuver pre-matching is specifically: first, a threshold is determinedAnd threshold->. Then, according to the average value of the absolute value of the first-order difference of the height of the subsequence to be identified, the average value is larger than a threshold value +.>Is divided into lifting-type actions, less than the threshold value +.>The sequence is divided into non-lifting actions. In the lifting action, according to the average value of the absolute value of the first-order difference of the pitch angle of the subsequence to be identified, the average value is larger than a threshold valueThe sequence of (1) is divided into a bucket-like body and a diving jump-like body, which is smaller than a threshold value +.>The sequence of (1) is divided into lifting turns; similarly, in the non-lifting type action, according to the average value of the absolute value of the first-order difference of the pitch angle of the sub-sequence to be identified, the average value is larger than a threshold value +.>Is divided into classes of convolutions, the mean value is smaller than the threshold value +.>Is divided into a sharp turn-like and a roll-like sequence. And removing maneuver templates which obviously do not accord with the sub-sequence category to be identified through maneuver pre-matching, so that the number of matched maneuver templates is reduced, the calculated amount is simplified, and the working efficiency is improved.
And S5, in a maneuver identification stage, similarity matching is carried out on the flight parameter subsequence to be identified and the target maneuver template sequence by using a multidimensional dynamic time planning method, so that the maneuver is finely classified.
Specifically, the maneuver identification stage is to perform similarity matching on the to-be-identified flight parameter subsequence obtained after S4 pre-classification and the existing target maneuver template sequences, and obtain a similarity value sequence by respectively calculating multidimensional dynamic time planning distances between the to-be-identified flight parameter subsequence and the C target maneuver template sequences,A value less than threshold->The action to be recognized is determined as +.>A corresponding standard action category. Wherein->Based on experience and man-madeGiven.
And the calculation of the similarity values of the two action sequences adopts the calculation principle of multidimensional dynamic time planning distance. Assume that action sequence 1 isAction sequence 2 is->MDTW path matrix is +.>。/>Representing the dimension of the action sequence, +.>And->Indicates the length of action sequence 1 and action sequence 2, respectively,/->Representing the weights of the respective dimensions.
The action sequences 1 and 2 are defined as:
wherein (1)>Is the +.1 of action sequence>Dimension features at%>The value of the individual points,/->Is the +.2 of action sequence>Dimension features at%>The value of each point.
The MDTW path matrix is defined as: ，/>wherein->Is +.1 of action sequence>Dimension features at%>The value of the individual points,/->Is +.2 for action sequence>Dimension features at%>The value of the individual points,/->Representation->Weight of dimension->Representing action sequence 1->Value vector of all dimension characteristics of each point and action sequence 2 +.>Frame weighted matching distance of value vectors of all dimension characteristics of each point;
s5-3, multidimensional dynamic time planning distance of action sequence 1 and action sequence 2Defined as the optimal regular path cumulative distance +.>And the regular path meets three constraint conditions of boundary, continuity and monotonicity, and the formula is defined as follows:
wherein the boundary condition is that the start point of the regular path must be +.>The termination point must be +.>The method comprises the steps of carrying out a first treatment on the surface of the The continuity condition is that the regular path should be continuous, i.e. jump from one point to the next, either right, up or up right; the monotonicity condition is that the direction of movement of the regular path should be monotonic, i.e. not move in the opposite direction. Constructing an optimal regular path according to the three constraint conditions, and calculating the accumulated distance of the optimal regular path to obtain +.>。
And S6, carrying out three-dimensional visualization on the sub-sequences of the flight parameters subjected to the fine classification, and carrying out manual accurate rechecking on the automatically marked maneuver class labels to obtain a final marked maneuver data set.
Specifically, the three-dimensional visualization of the flight attitude and the track is carried out on the flight parameter subsequence automatically and finely classified by the multidimensional dynamic time planning method, and the manual accurate rechecking is carried out on the automatically marked maneuver type label. When the classification result of the multi-dimensional dynamic time planning method is consistent with the manual classification result, the flight parameter subsequence and the corresponding label category are stored, and when the classification result of the multi-dimensional dynamic time planning method is inconsistent with the manual classification result, the manual classification result is used as a category label, and the corresponding flight parameter subsequence is stored, so that the labeled maneuvering action data set is obtained. In summary, in a preferred embodiment provided in the present application, a detailed implementation flowchart of automatic auxiliary labeling of a maneuver data set is shown in fig. 4.
Referring to fig. 5, an automatic auxiliary labeling device 100 for a complex maneuver data set of a military machine according to an embodiment of the present application includes:
a data acquisition module 11, configured to acquire flight parameter sequence data;
the preprocessing module 12 is configured to perform data preprocessing on the flight parameter sequence data to generate preprocessed flight parameter sequence data;
a pre-matching module 13 for performing preliminary matching on the target action template and the historical flight parameter sequence to obtain a sub-sequence of flight parameters to be identified
And the maneuver classification module 14 is configured to perform similarity matching calculation on the sub-sequence of the flight parameter to be identified and the target maneuver template sequence, so as to obtain a fine classification result of maneuver.
The maneuver three-dimensional visualization module 15 is used for performing three-dimensional visualization of the flight gesture and the track on the maneuver sequences subjected to the fine classification, and further performing manual accurate review on the maneuver types automatically marked.
The invention provides an automatic auxiliary labeling method for a complex maneuver data set of a military machine type by referring to the thought of multi-dimensional time subsequence matching query, and aims to automatically detect and label a target maneuver from historical flight data. Firstly, adopting a Matrix Profile data structure, and preliminarily matching and extracting a target maneuver subsequence in mass data by using a maneuver template; secondly, determining the maneuver type of the segmented maneuver sub-sequence by adopting a multidimensional dynamic time planning (MultidimensionalDynamic Time Warping, MDTW) method; and finally, carrying out three-dimensional visualization on the flight gesture and the track on the maneuvering action sequence so as to manually and accurately recheck the maneuvering type automatically marked. The automatic auxiliary labeling method can remarkably improve the labeling efficiency of the complex maneuvering action data set, is suitable for various military machine types, and has wide engineering application prospect.
The foregoing is merely exemplary of the present application and is not intended to limit the present application. Various modifications and changes may be made to the present application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc. which are within the spirit and principles of the present application are intended to be included within the scope of the claims of the present application.
Claims (4)
1. The automatic auxiliary labeling method for the complex maneuvering data set of the military machine type is characterized by comprising the following steps of:
s1, acquiring historical flight parameter sequence data of pilot flight training, and selecting a standard maneuver sample sequence as a target maneuver template;
s2, preprocessing the collected historical flight parameter sequence data and the selected target maneuver template sequence;
s3, based on a Matrix Profile data structure, using a target maneuvering template sequence to preliminarily match in a historical flight parameter sequence subjected to data preprocessing and extracting a flight parameter subsequence to be identified;
s4, pre-classifying the sub-sequences of the flight parameters to be identified extracted in the S3 through obvious features, namely pre-matching the maneuvering types;
s5, performing similarity matching on the to-be-identified flight parameter subsequence and the target maneuver template sequence in the pre-classified maneuver sequence by adopting an MDTW algorithm, and completing fine classification of maneuver;
s6, carrying out three-dimensional visualization on the sub-sequences of the flight parameters subjected to fine classification, and rechecking the automatically marked maneuver class labels to obtain a final marked maneuver data set;
in step S3, based on the Matrix Profile data structure, a target maneuver template sequence is used to preliminarily match and extract a subsequence of a flight parameter to be identified in a historical flight parameter sequence subjected to data preprocessing, which specifically includes:
given a time series of data-preprocessed historical flight parametersTarget maneuver template sequence for querying +.>Sliding from the initial position of the historical flight parameter time sequence F subjected to data preprocessing by using a sliding window with the length of the target maneuver template sequence Q, calculating the distance between a subsequence in the window and the target maneuver template sequence Q each time, generating Matrix Profile with the length of n-m+1, and inquiring a value smaller than a threshold value gamma in the Matrix Profile, wherein the position of the value is the subsequence of the flight parameter to be identified, which is matched and extracted from F>
Where k represents the number of parameters, n represents the length of the time series of historical flight parameters, m represents the length of the sequence of target maneuver templates, and typically m is much smaller than n;
in step S6, the sub-sequence of the flight parameters after fine classification is visualized in three dimensions, and the automatically labeled maneuver class labels are manually and precisely checked to obtain a final labeled maneuver data set, which specifically includes:
carrying out three-dimensional visualization on flight parameter subsequences automatically and finely classified by an MDTW algorithm by using pitch angle, roll angle and yaw angle information, carrying out three-dimensional visualization on flight tracks by using altitude, longitude and latitude information on the flight parameter subsequences automatically and finely classified by the MDTW algorithm, rechecking automatically marked maneuver type labels according to three-dimensional visualization results of the flight gestures and the tracks, storing the flight parameter subsequences and corresponding label types when the classification result of the MDTW algorithm is consistent with the manual classification result, taking the manual classification result as a category label when the classification result of the MDTW algorithm is inconsistent with the manual classification result, and storing the corresponding maneuver parameter subsequences, so as to obtain maneuver action data sets with labels;
in step S4, the pre-classifying the sub-sequence of the flight parameter to be identified extracted in step S3 by using the obvious features, that is, performing pre-matching of maneuver types, specifically includes:
s4-1, the obvious characteristics are height and pitch angle, and the pre-classified categories comprise a bucket like a jin, a jump like a dive, a lift turn like a sharp turn, a roll like a roll and a spiral like four categories;
s4-2, pre-matching of the maneuvering action category is specifically as follows: firstly, setting a threshold H and a threshold P, wherein the threshold H is an average value of the maximum height difference value and the minimum height difference value in each standard target maneuvering template sequence, the average value of the maximum pitch angle difference value and the minimum pitch angle difference value in each standard target maneuvering template sequence is obtained by P, then, according to the average value of the first-order difference absolute values of the heights of the subsequences of the flight parameters to be identified, the sequence with the average value larger than the threshold H is divided into lifting type actions, the sequence with the average value smaller than the threshold H is divided into non-lifting type actions, in the lifting type actions, according to the average value of the first-order difference absolute values of the pitch angles of the subsequences of the flight parameters to be identified, the sequence with the average value larger than the threshold P is divided into a bucket like and a dive ascent like, and the sequence with the average value smaller than the threshold P is divided into lifting type turns; in the non-lifting type actions, according to the average value of the pitch angle first-order difference absolute values of the subsequences to be identified, the sequence with the average value larger than the threshold value P is classified into a class spiral, and the sequence with the average value smaller than the threshold value P is classified into a class sharp turn and a class roll.
2. The method for automatically assisting in labeling a complex maneuver data set for a military machine according to claim 1, wherein in step S1, historical flight parameter sequence data of pilot flight training is collected, and a standard maneuver sample sequence is selected as a target maneuver template, specifically comprising:
s1-1, acquiring historical flight parameter sequence data of pilot flight training, and decoding the historical flight parameter data into table data which can be directly read by a computer;
s1-2, determining the type of the target maneuver to be marked according to the flight training outline and the actual demand, and selecting a flight parameter subsequence meeting maneuver operation specification requirements in the flight training manual as a standard target maneuver template sequence.
3. The method for automatically assisting in labeling a complex maneuver data set for a military machine according to claim 1, wherein in step S2, the collected historical flight parameter sequence and the selected target maneuver template sequence are subjected to data preprocessing, and specifically comprises:
and filling missing values in the historical flight parameter sequence and the target maneuver template sequence by taking parameters of pitch angle, inclination angle, barometric altitude, X-axis angular velocity, Y-axis angular velocity, Z-axis angular velocity, horizontal acceleration and vertical acceleration which can be judged in different action types, and carrying out standardized preprocessing on each dimension in the original data.
4. The automatic auxiliary labeling method for complex maneuver data sets of military machine type according to claim 1, wherein in step S5, similarity matching is performed between the sub-sequence of the flight parameter to be identified and the target maneuver template sequence by using an MDTW algorithm, so as to complete fine classification of maneuver, and specifically comprises:
s5-1, performing similarity matching on the to-be-identified flight parameter subsequence obtained after S4 pre-classification and the existing target maneuver template sequence, and respectively calculating MDTW distances between the to-be-identified flight parameter subsequence and the c target maneuver template sequences to obtain a similarity value sequence S= [ S ] 1 ,...,S c ]，S i A value less than the threshold τ i ∈τ＝[τ 1 ,...,τ c ]Then will be identifiedIs determined as S i Corresponding standard action categories, wherein tau is set according to MDTW distance values of the actual maneuver sequence and the corresponding target maneuver template sequence;
s5-2, calculating similarity values of two action sequences by adopting the MDTW distance calculating principle, and assuming the action sequence 1 isAction sequence 2 is +.>MDTW path matrix is P, k represents the dimension of the action sequence, m and n represent the lengths of action sequence 1 and action sequence 2, a [1, …, k, respectively]Weights representing the respective dimensions;
the action sequences 1 and 2 are defined as:
wherein x is km The value of the kth dimension characteristic of the action sequence 1 at the mth point is y kn The value of the kth dimension characteristic of the action sequence 2 at the nth point is taken;
the MDTW path matrix is defined as:
wherein x is qi The q dimension feature of the action sequence 1 is valued at the ith point, y qj The q dimension feature of the action sequence 2 is valued at the j point, a q Weights representing the q dimension, e ij The frame weight matching distance of the value vectors of all the dimension characteristics of the ith point of the action sequence 1 and the value vectors of all the dimension characteristics of the jth point of the action sequence 2 is represented;
s5-3, the multidimensional dynamic time planning distance MDTW (X, Y) of the action sequence 1 and the action sequence 2 is defined as the optimal regular path accumulated distance r ij And the regular path meets three constraint conditions of boundary, continuity and monotonicity, and the formula is defined as follows:
MDTW(X,Y)＝min{r mn }
r ij ＝e ij +min{r (i-1)(j-1) ,r (i-1)j ,r i(j-1) },0<i≤m,0<j≤n,
wherein the boundary condition is that the starting point of the regular path must be e 11 The termination point must be e mn The method comprises the steps of carrying out a first treatment on the surface of the The continuity condition is that the regular path should be continuous, i.e. jump from one point to the next, either right, up or up right; the monotonicity condition is that the moving direction of the regular path should be monotonic, namely, the regular path will not move reversely, an optimal regular path is constructed according to the three constraint conditions, and the accumulated distance of the optimal regular path is calculated to obtain r ij 。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311116892.7A CN116861300B (en) | 2023-09-01 | 2023-09-01 | Automatic auxiliary labeling method and device for complex maneuvering data set of military machine type |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311116892.7A CN116861300B (en) | 2023-09-01 | 2023-09-01 | Automatic auxiliary labeling method and device for complex maneuvering data set of military machine type |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116861300A CN116861300A (en) | 2023-10-10 |
| CN116861300B true CN116861300B (en) | 2024-01-09 |
Family
ID=88221860
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311116892.7A Active CN116861300B (en) | 2023-09-01 | 2023-09-01 | Automatic auxiliary labeling method and device for complex maneuvering data set of military machine type |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116861300B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106092093A (en) * | 2016-05-26 | 2016-11-09 | 浙江工业大学 | A kind of indoor orientation method based on earth magnetism fingerprint matching algorithm |
| CN109241077A (en) * | 2018-08-30 | 2019-01-18 | 东北大学 | Production target variation tendency visual query system and method based on similitude |
| CN111390921A (en) * | 2020-03-16 | 2020-07-10 | 华东师范大学 | Library uses multi-functional robot |
| WO2022068142A1 (en) * | 2020-09-30 | 2022-04-07 | 华人运通(上海)云计算科技有限公司 | Road surface terrain recognition method and apparatus based on suspension vibration signal, and storage medium |
| CN114299422A (en) * | 2021-12-20 | 2022-04-08 | 中国人民解放军海军航空大学 | Flight quality self-adaptive evaluation method |
| CN115169403A (en) * | 2022-07-11 | 2022-10-11 | 中联重科股份有限公司 | Method and device for identifying operation mode of mechanical equipment |
| CN115561575A (en) * | 2022-09-09 | 2023-01-03 | 国网上海市电力公司 | Method for distinguishing electrical abnormal state of offshore wind farm based on multi-dimensional matrix profile |
| CN116010889A (en) * | 2022-12-30 | 2023-04-25 | 中国人民解放军海军航空大学 | Intelligent recognition method for abnormal flight state of aviation aircraft |
-
2023
- 2023-09-01 CN CN202311116892.7A patent/CN116861300B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106092093A (en) * | 2016-05-26 | 2016-11-09 | 浙江工业大学 | A kind of indoor orientation method based on earth magnetism fingerprint matching algorithm |
| CN109241077A (en) * | 2018-08-30 | 2019-01-18 | 东北大学 | Production target variation tendency visual query system and method based on similitude |
| CN111390921A (en) * | 2020-03-16 | 2020-07-10 | 华东师范大学 | Library uses multi-functional robot |
| WO2022068142A1 (en) * | 2020-09-30 | 2022-04-07 | 华人运通(上海)云计算科技有限公司 | Road surface terrain recognition method and apparatus based on suspension vibration signal, and storage medium |
| CN114299422A (en) * | 2021-12-20 | 2022-04-08 | 中国人民解放军海军航空大学 | Flight quality self-adaptive evaluation method |
| CN115169403A (en) * | 2022-07-11 | 2022-10-11 | 中联重科股份有限公司 | Method and device for identifying operation mode of mechanical equipment |
| CN115561575A (en) * | 2022-09-09 | 2023-01-03 | 国网上海市电力公司 | Method for distinguishing electrical abnormal state of offshore wind farm based on multi-dimensional matrix profile |
| CN116010889A (en) * | 2022-12-30 | 2023-04-25 | 中国人民解放军海军航空大学 | Intelligent recognition method for abnormal flight state of aviation aircraft |
Non-Patent Citations (6)
| Title |
|---|
| Time series motifs discovery under DTW allows more robust discovery of conserved structure;Sara Alaee等;Data Mining and Knowledge Discovery;第35卷;863–910 * |
| 军用机型复杂飞行动作提取与识别的方法研究;李超等;航空兵器;第30卷(第1期);第2节 * |
| 基于MDTW的飞行动作识别算法;李鸿利等;计算机工程与应用;第51卷(第9期);第2.2节 * |
| 基于多元时间序列融合的飞行动作识别方法;周超等;空军工程大学学报(自然科学版);第18卷(第4期);摘要，第2节，图3 * |
| 快速时间序列模体发现算法;朱晓晓;;国外电子测量技术(09);31-35 * |
| 时间序列可变尺度的时频特征求解及其分类;魏池璇等;软件学报;第33卷(第12期);4411-4428 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116861300A (en) | 2023-10-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111062282B (en) | Substation pointer instrument identification method based on improved YOLOV3 model | |
| CN110070074B (en) | Method for constructing pedestrian detection model | |
| CN108090499B (en) | Data active labeling method and system based on maximum information triple screening network | |
| CN111414932B (en) | Classification identification and fault detection method for multi-scale signals of aircraft | |
| CN110798805B (en) | Data processing method and device based on GPS track and storage medium | |
| CN105654139A (en) | Real-time online multi-target tracking method adopting temporal dynamic appearance model | |
| CN112859898B (en) | Aircraft trajectory prediction method based on two-channel bidirectional neural network | |
| CN107798691B (en) | A kind of unmanned plane independent landing terrestrial reference real-time detection tracking of view-based access control model | |
| CN109159113A (en) | A kind of robot manipulating task method of view-based access control model reasoning | |
| CN111652835A (en) | Method for detecting insulator loss of power transmission line based on deep learning and clustering | |
| CN110490166A (en) | High iron catenary pipe cap automatic testing method based on depth learning technology | |
| CN111368762A (en) | Robot gesture recognition method based on improved K-means clustering algorithm | |
| CN114298183B (en) | Intelligent recognition method for flight actions | |
| CN117574283A (en) | Complex flight action segmentation and detection method and device based on flight parameters | |
| CN114359604B (en) | Formation recognition model construction method based on multi-aircraft trajectory data | |
| CN116861300B (en) | Automatic auxiliary labeling method and device for complex maneuvering data set of military machine type | |
| CN117288094B (en) | Real-time positioning system of heading machine based on laser sensor | |
| CN115019294A (en) | Pointer instrument reading identification method and system | |
| CN114662594A (en) | Target feature recognition analysis system | |
| CN111241905A (en) | Power transmission line nest detection method based on improved SSD algorithm | |
| CN109753922A (en) | Anthropomorphic robot expression recognition method based on dense convolutional neural networks | |
| CN114463425A (en) | Workpiece surface featureless point positioning method based on probability Hough linear detection | |
| CN118094213A (en) | Deep learning-based aircraft maneuver identification system | |
| Wang et al. | Object detection algorithm based on improved Yolov3-tiny network in traffic scenes | |
| CN112084353A (en) | Bag-of-words model method for rapid landmark-convolution feature matching |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |